Research on Space Vector PWM Inverter Based on Artificial Neural Network

Abstract

This paper proposed a Space Vector PWM algorithm based on artificial neural network for voltage-source inverters. When calculating the invert's three-phase turn-on time, the paper uses a three-layer forward-feed network which adopts the algorithm of Levenberg-Marquarde to train the network. This method uses artificial neural network's strong nonlinear approximation ability to avoid a lot of nonlinear calculation. At last, in the environment of MATLAB/Simulink, simulation model of the system was built. The simulation results show that the SVPWM algorithm of artificial neural network can improve the switching frequency and reduce the harmonic of output voltage and current.